History has repeatedly shown us isolated examples of completely replacing the need for human labour.
The industrial revolution did destroy jobs, but back then labour was incomprehensibly unproductive. So much so that around 200 years ago 50% of the entire workforce worked in agriculture.
This meant that although a large percentage of the total workforce was affected by the industrial revolution the total percentage of jobs that machines could do was absolutely tiny. Even where machines could typically only automate some percentage of the total work. For example, in agriculture although machines could do a lot of the work, there was still a need for humans to operate the machines.
However, there were some exceptions. Texture weavers were completely replaced by machines – it didn't make them more efficient it made them unemployed. More recently human calculators have been completely replaced. And more recently still checkout staff at supermarkets are being replaced by self-service systems. Again, a self-checkout machine doesn't make a job more productive, it entirely replaces the need for a human to do it.
Even today there's still relatively very few jobs which can be entirely done by machines, but for the first time ever, we're starting to see how this could change.
Then the question then is what new jobs might be created if you no longer need graphic designers or VFX artists or lawyers? If humans aren't good for physical labour or mental labour then what are they good for?
The assumption that humans will remain employable assumes that humans will still be able to provide some economic value which machines cannot. That humans unlike horses won't be sent to the glue factory because there will still be some economic use for us so that we can still be net providers and not a net-burden.
Even if we assume there will be a small percentage of jobs which machines won't be able to do, if you have the entire workforce competing for those jobs then you'll probably be paid so poorly you might as well not have a job.
Would be interested what you disagree with. Is it that you don't believe that AI will be able to replace humans entirely? If so what jobs do you think are safe? Or is it that you believe new jobs will be created? If so can you describe what these jobs might involve that machines cannot do?
I've been actively prepping for very bad economic and political scenarios, plus some doom scenarios for about 5 years now.
My assumption is that white collar unemployment will become a serious problem by the end of the decade, but potentially as soon as 2-3 years. I believe ASI will come within 10 years and it's hard to predict past the intelligence singularity, but my assumption would be that rapid ASI-assisted robotic advancements will make the majority of humans in developed nations unemployed within a two decades. But again, I can't really predict what will happen post-ASI. I could be very wrong if just a few variables change.
The unemployment concern is the bit I'm most certain of but that also worries me the least personally, since it's by far the easiest risk to prep for. If I'm made unemployed and never find work again I'll be fine.
I'm also an accelerationist when it comes to AI-driven unemployment because I think an crisis-level unemployment spike before ASI arrives is probably our only hope at getting off the AI doom train we're on before it's too late.
However, more recently I have been having fun by having three concurrent projects going at once. Instead of having less work to do, I just broadened the scope of what I was going to do.
Suddenly, it became so much more interesting. I have started multiple porting-to-WASM projects, a Jellyfin clone that I am already running on my home server written in Erlang, new themes for my blog, and many other things.
I realized that over the years there has been dozens (hundreds?) of projects that I have wanted to do, but I never really got around to doing them because it was just too much effort and I couldn’t justify the time sink. By having multiple agents working at once, I can work on multiple projects concurrently, and I can focus on the harder (and more fun) parts of programming.
Do you feel like porting projects from one language to another is actually that productive?
Like, that's fine if you are having fun, but that has no bearing on this overall discussion about automating all the useful work so that we are no longer needed
My personal projects don’t get paid, no, they’re just for fun, but I feel like my point was missed: instead of “having less to do”, I broadened the scope.
I feel like people are making an incorrect assumption that “labor”, even white collar labor, is a fixed-size problem. I don’t agree with that: there’s never going to be a point where we say “shit, there is literally nothing left to do”.
To me looks like we aren't and your personal projects don't support the opposite.
I use Codex for work, and I feel like I accomplish more with it than I would have without it. I juggle a lot more projects concurrently and can broaden the scope of what I am doing.
Now we can argue what I do is “easy”, and you’re probably not wrong, but most white collar work is pretty easy and being able to accomplish more easy tasks is still potential productivity.
For that matter, if I am not distracted with the trivial tasks I can dedicate more of my time working on hard problems, maybe even ones we would call “important”.
Tangential, but related.
When I first started my career, it was working for a company that used SVN as its version control (Git hadn’t quite won yet). Every Friday, we would do The Merge. We called it that, like a proper noun.
The Merge was almost always a 4-6 hour process because it would be taking every branch of every feature/bug we have worked on, and merged it into trunk. The day would be then spent fixing merge conflicts as they came up. I was the most junior engineer (borderline intern) and so that task generally fell on me.
Eventually, I found git-svn, and it brought a multi-hour task down to like 30 minutes, since the merging was so much quicker.
Fridays were always blocked out for me to do The Merge, but it’s not like when I got it done in 30 minutes they said “oh, umm, there’s nothing else to do, go home and forget being paid for today”. They just found other stuff for me to do and I did that.
There’s always more stuff that can be done. We won’t ever run out of potential shit we can do.
One thing I have seen no evidence of is AI being smarter than the smartest humans. It's still just a search engine (with an impressive human language UI) of existing human knowledge. It cannot do anything it has not been trained on.
LLM: crawl the entire Internet and train a model. When you ask it a question, it uses that model to spit out an answer. Worse than a search engine though, if it can't come up with a good answer it just makes shit up
Both use the same sources of information. Neither can produce a (correct) answer that doesn't exist somewhere in that information source
Because this kind of talk always sounds like it’s because people thought The Terminator was a documentary.
Write a book about it then?
Why? Idk